Date: 24 Apr 2002 23:59:36 +1000 From: Mark Sergeant <msergeant@looksmart.net> To: bv@wjv.com Cc: freebsd-isp@FreeBSD.ORG Subject: Re: HTTP Load Balancing and Availability Solutions Message-ID: <1019656776.38204.17.camel@xyzzy.intranet.snsonline.net> In-Reply-To: <20020424132125.GE85736@wjv.com> References: <3CC6A5D2.3070701@emre.de> <1019652482.38204.4.camel@xyzzy.intranet.snsonline.net> <20020424132125.GE85736@wjv.com>
next in thread | previous in thread | raw e-mail | index | archive | help
99.999 is less than an hours down time per year which running off one web server is easily achievable, even with monitoring, replacement hardware etc on hand it will still take a while @ 3am when something goes wrong for a person to be paged, getup, ascertain problem, pull in redundant hardware to do the web serving etc. This also doesn't take into account a monthly patch schedule for whichever OS running whichever software. Other things to take into account are network outages, mis-configurations / patches going wrong etc. When running a production environment with aggressive uptime requirements and also having things rolled out and changed on a daily basis the only true way to achieve the required uptime is via hardware load balancing and redundant links / redundant hardware. I run a web hosting company which guarantees 95% uptime to its clients for normal everyday web hosting which is plenty for most companies and enables me to take machines down every month for patches, upgrades and also covers for the other unfortunate times as I don't have access to the noc 24/7 without paying ridiculous call out fees. I also in my day job manage 3 web farms on 2 continents providing a 99.999% uptime guarantee and the only thing that lets me sleep at night is the fact hardware load balancing and redundant machines / load balancers are in place. Just my 0.02 cents :) On Wed, 2002-04-24 at 23:21, Bill Vermillion wrote: > On Wed, Apr 24, 2002 at 22:48 , while impersonating an expert on=20 > the internet, Mark Sergeant sent this to stdout: >=20 > > Depending on the nature of the project and the availability > > required I would only recommend a software based solution for a > > project that didn't have a requirement of more than 95% uptime, >=20 > 95% UPTIME. That is only 51 out of 52 weeks/year. Even 99.999% > is 8 hours down time per year. Five 9' is usually achievable with > just good hw/sw design. My web server - though not heavily > stressed is a 498 days 16:44 as I type this. >=20 > > My recommendation would be should it absolutely have to be done with > > software then go with the mod_proxy and a 1 hour ttl for your www > > address (anything less than an hour is usually ignored), this way a > > replacement box can take over the proxying server inside of an hour > > should there be any issues. >=20 > Would it not make sense to monitor the target machine and if/when > it fails then just re-write the IP on the standby machine to > achieve a fail-over scenario. >=20 > Bill > --=20 > Bill Vermillion - bv @ wjv . com >=20 > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-isp" in the body of the message >=20 --=20 Mark Sergeant Senior Unix Systems Administrator =20 L=F4=F4kSmart International Pty. Ltd. Level 5/388 Lonsdale Street Melbourne, VIC, 3000 Australia=20 P. (03) 9648 2201=20 F. (03) 9648 2244=20 http://www.looksmart.com.au The referring document contains privileged and confidential information. If you are not the intended recipient you must not copy, distribute or take action with regards to the content, we request that you notify LookSmart International Pty. Ltd. immediately and remove all traces of this document.=20 Any views expressed in this message are those of the individual sender, except where they are specifically stated to be the views of LookSmart=20 International Pty. Ltd. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-isp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1019656776.38204.17.camel>